Proposed LDAS validation experiments with Oklahoma Mesonet and ARM/CART observations
Objective
Comparison between LDAS forcing and the locally observed forcing shows a general agreement between these two datasets, as a result of
forcing validation. The discrepancy between the model generated hydrological variables,especially soil moisture and the observations still exists.
Now we are trying to run the models with the locally observed forcing and the local land surface characteristics to separate the uncertainties associated with the forcing and the models.
Experiment Setup
The following four runs will be carried out by all modeling group.
- EXP000 -------original LDAS run
The forcing files ARE THE ORIGINAL FULL-DOMAIN NASA FORCING.
The land mask is here . This is a binary file (464*224 pixles) with 1 indicating land and 0 indicating water.
This mask consists 72 oklahoma mesonet stations, 24 ARM/CART station, 19 Illinois stations and 14 HPRCC stations in Nebrask. You can use
the following GrADS control file to display the mask.
- EXP111 -------all available local forcing are used
The forcing files are available here , or
ftp climate.envsci.rutgers.edu as anonymous user
cd /pub/luo/EXP111
bin
prompt
mget *.tar
bye
This set of forcing includes all the possible locally observed variables. It is in exactly the same format as original LDAS forcing, but the following variables are substituted by local forcing.
TMP ...............................USE LOCAL TEMPERATURE
SPFH ...............................USE LOCAL SPECIFIC HUMIDITY
PRES ...............................USE LOCAL PRESSURE
UGRD ...............................USE LOCAL UGRD
VGRD ...............................USE LOCAL VGRD
ETA DSWRF
DLWRF
ETA APCP
ETA ACPCP
DSWRF...............................USE LOCAL SHORTWAVE
BRTTMP
PAR
APCP ...............................USE LOCAL PRECIP instead of APCP
STAGEIV APCP
You basically do not have to change anything when using this forcing. If the local PRECIP or SHORTWAVE is not available, you should use ETA PRECIP and ETA shortwave, which is the standard way in original LDAS rerun.
This different land mask is used in this run.
This is a binary file (464*224 pixles) with 1 indicating land and 0 indicating water.
This mask consists 72 oklahoma mesonet stations, 13 ARM/CART station. Because We only have local forcings over these stations. You can use
the following GrADS control file to display the mask.
EXP111-2 -------EXP111 forcing with local soil texture map and Cosby/Rawl parameter table
Using the EXP111 forcing and one of the two soil map: multilayer or 1 layer . This soil map is derived from the local observations at those stations. It is probably not the most predominant soil type in the grid, but the soil type at the point where soil moisture are observed.
The file gives the OK Mesonet station id or ARM/CART station id, lat and lon, x and y in the standard LDAS grid, soil type in 4 layers or intergrated to 1 layer using the program Matthew Rodell provided. (Thanks!).
VIC and NOAH model can only use the intergrad 1 layer soil texture, while MOSAIC can use the multilayer one. Because we don't have all the information, we can only provide soil information for 73 stations out of 85. Other station you should use the
the original soil texture you have for rerun. If it is a big problem for you, let me know.
Using the following FORTRAN code to read the file:
character(len=4) :: siteid
real :: lat, lon
integer, dimension(4) :: tex
integer :: x, y
read(*,300)siteid,lat, lon, x, y, (tex(j),j=1,4)
300 format(a4,3x,f9.4,3x,f9.4,3x,i4,3x,i4,3x,i2,3x,i2,3x,i2,3x,i2)
Here is the soil type code:
TEX = 1 !CLAY
TEX = 2 !SILTY CLAY
TEX = 3 !SANDY CLAY
TEX = 4 !CLAY LOAM
TEX = 5 !SILTY CLAY LOAM
TEX = 6 !SANDY CLAY LOAM
TEX = 7 !LOAM
TEX = 8 !SILTY LOAM (and SILT)
TEX = 9 !SANDY LOAM
TEX = 10 !LOAMY SAND
TEX = 11 !SAND
Here is the message from Ken Mitchell regarding the experiment:
Brian, Justin, Eric, Dag, Qingyun, Lifeng, Alan,
Recall from our last N-LDAS phone conferences of 11 and 18 April, we
agreed to make another reduced-grid retro N-LDAS run, designated as Case
"111-2" (using OU Mesonet observed local forcing from Lifeng Luo on the
"OU" 84-pixel reduced grid), in which the "-2" of "111-2" designates our
use of:
A) common "locally observed" soil texture class index at each of the 84
pixels (the index range being 1-12 corresponding to the 12 soil class
indices of Cosby et al.)
We are to use Lifeng Luo's "map" of the soil texture class index at the
84-pixels (this map is based on the OU Mesonet-folk determined locally
sampled soil texture).
B) common 12-class table of FOUR soil physical parameters
1 - porosity
2 - B-parameter
3 - saturated soil matric potential
4 - saturated soil hydraulic conductivity
In response to Brian's email below, clarifying what he used in MOSAIC, I
direct that in this Experiment 111-2 that all four models (MOSAIC, NOAH,
VIC, SAC) use the "Cosby/Rawls" soil parameter table from Qingyun Duan,
given below by Brian, namely the " Cosby parameters for everything
except for Ksat from Rawls" (WITHOUT substituting Rawls values for the
sand class, except for Ksat of course).
BUT NOTE AGAIN: INVOKE TABLE BELOW ONLY FOR THE FOUR SOIL PARAMETERS
LISTED ABOVE UNDER "B". Otherwise continue to use what your model has
been using, except of course for those parameters that your model
calculates from the above four.
CAUTION: double check that you adhere to your model's units convention
and sign convention for these four soil parameters (e.g. in some models
the saturated soil matric potential should be given as a negative value
if I recall).
ALSO: In a closely related follow-on experiment, I propose we also
execute an Experiment "111-3", in which in addition to the above common
four soil parameters and common soil type map, we also commonly adopt
the following:
1 - saturated soil thermal conductivity from the table below
2 - field capacity from the table below
3 - wilting point from the table below
4 - common universal vegetation class of "grass cover" at each of the 84
pixels
5 - common vegetation parameters for "grass cover" to the extent that we
can (e.g. minimal stomatal resistance, seasonal cycle of fractional area
of green vegetation cover, etc -- the source and value of these will
have to be discussed at our next N-LDAS phone conference of April 26, 10
AM).
Ken Mitchell
Cosby+Rawls
0.339 0.0854 0.0218 0.0692 3.7472E-05 2.79 0.92
0.421 0.1467 0.0599 0.0363 1.8056E-05 4.26 0.82
0.434 0.2242 0.1002 0.1413 6.3611E-06 4.74 0.60
0.476 0.3626 0.1772 0.7586 2.1389E-06 5.33 0.25
0.476 0.3626 0.1772 0.7586 2.1389E-06 5.33 0.10
0.439 0.2881 0.1393 0.3548 1.4444E-06 5.25 0.40
0.404 0.2527 0.1438 0.1349 9.7222E-07 6.77 0.60
0.464 0.3837 0.2477 0.6166 1.1944E-06 8.72 0.10
0.465 0.3420 0.2144 0.2630 3.3333E-07 8.17 0.35
0.406 0.2930 0.2053 0.0977 2.7778E-07 10.73 0.52
0.468 0.3750 0.2597 0.3236 4.4444E-07 10.39 0.10
0.457 0.3117 0.2240 0.4677 5.2778E-07 11.55 0.25
The soil classes (rows):
1 S Sand
2 LS Loamy sand
3 SL Sandy loam
4 SIL Silt loam
5 SI Silt
6 L Loam
7 SCL Sandy clay loam
8 SICL Silty clay loam
9 CL Clay loam
10 SC Sandy clay
11 SIC Silty clay
12 C Clay
soil property definition (Col.):
1: Porosity
2: Field capacity
3: Wilting point
4: Saturated soil matric potential (in m of h2o)
5: Saturated soil hydraulic conductivity (in m/s)
6: B parameter
7: Quatz content
Inital condition
Modeling Period
Submitting Results
Please finish these runs AS SOON AS POSSIBLE. VIC output can be stored somewhere on water.princeton.edu since I have access to this machines. MOSAIC output can be stored on
land.gsfc.nasa.gov. NOAH model output can be ftped to Rutgers machine through existing account. Just submit the output in grib format, no script is needed to extract the data. We will take care of it.
